INFLUENCE OF CERTAIN ELEMENTS ON STEEL. 349 be incorrectly ascribed to whatever small percentage of silicon has .survived the reactions during recarburization. This criticism on the determination of carbon applies to the data given in Tables XVII-A and XVII-C, and renders the calculations thereon of limited value. Many continental works have habitually made rails with from .30 to .60 per cent, of silicon, and all requirements of strength and ductility have been met. All the authorities do not approve this practice, and it is stated by Ehrenwerth* that the latest results are rather in the opposite direction in the case of low steels, f but I was told some years ago, by the manager of one of the French establishments, that the only way in which he was able to fill one contract with particularly severe specifications was by making the rails contain from .30 to .40 per cent, of silicon, since a less proportion would not stand the drop-tests. It is not necessary to question whether this conclusion was warranted or not; it is enough, to know that the steel was of the best quality, whether on account of the silicon or in spite of it. Silicon is allowed in rails by Sahdberg, who writes as follows :J ."Silicon up to .30 per cent., with carbon .30 to .40 per cent., does not harden steel or make it brittle, and diminishes its strength in such small degree as not to imperil the safety of the rail." The italics are my own, and call attention to the implication that silicon lowers the strength rather than raises it. Exceptional cases have been recorded of soft steels with high silicon, like the tough rail mentioned by Snelus,§ with carbon below .10 per cent, and .silicon .83 per cent. It must be considered, however, that although this might have been very tough for a rail, it does not follow that it was very tough for soft steel, but it is quite certain that it could not have been bad or brittle. Knowing the relative effect of impurities upon hard and soft steels, the assumption would be justified that low-carbon metal could contain a larger percentage of silicon than higher steel, but • structural steels do not often contain over .05 per cent, of silicon, while usually they hold less than .03 per cent. Tool steel is sub- * Das Berg- und Hilttenweaen auf der Weltausstellung in Chicago, 1895. t See page 78, ante. t Proc. English Inst. Mech. Sng., 1890, p. 301. § On the Chemical Composition and Tetting of Steel Bails, Journal I. and S. I., Vol. II, 1882, p. 583.